Four Independent Prediction Engines + Signal Confluence

• APY Momentum Engine (V3.0): Detects statistically significant yield deviations using Z-score normalisation (≥ 2.0σ), RSI confirmation, Bollinger Band position, and multi-timeframe EMA alignment (3/7/14/21-day). Primarily predicts "Cooling Down" signals. Heating-up predictions require extreme conditions (Z ≥ 3.0σ with full indicator alignment). Pools with ≤ 25% accuracy after 5+ predictions are automatically blacklisted. Verification window: 3–7 days (adaptive).
• Yield Reliability Index (V5.0): Measures yield consistency against a 5% APY benchmark using a composite quality score: reliability (50% weight), stability (30%), and current yield (20%). Predicts whether pools will maintain benchmark performance or decline. Only pools with 60%+ success rate, $500k+ TVL, and 20+ days of data qualify. Verification window: 14 days.
• TVL Momentum Engine (V4.0): Tracks capital flows using EMA(3) vs EMA(14) crossovers on Total Value Locked data. Predicts stabilisation only — directional predictions (capital inflow at 29% accuracy, capital flight at 37%) have been retired as they performed worse than a coinflip. Signals for inflow and outflow are still displayed for informational context but do not generate scored predictions. Verification window: 4 days.
• Surge Predictor (V3.0): Focuses exclusively on extreme utilisation scenarios (96%+) where algorithmic interest rate models force rate spikes. This is the highest-accuracy engine — extreme surge predictions historically achieve 84%+ accuracy. All other utilisation prediction types (Stable, Decay, Rising) have been permanently retired. Verification window: 48 hours.
• Signal Confluence: Combines all four engine scores (0–25 points each) into a unified signal strength (0–100). When multiple engines agree on a pool's direction, confidence increases. Signals are verified after 48 hours against APY and TVL movements.

Verified Performance & Transparency

Every prediction is publicly auditable with timestamped outcomes:

• Rolling Accuracy Windows: Every engine page displays 7-day, 30-day, and all-time accuracy — so users can see whether performance is improving or degrading
• NEUTRAL Exclusion: Only definitive correct/incorrect verdicts count in accuracy calculations. Ambiguous outcomes from legacy predictions are excluded from the denominator for honest reporting
• Prediction Type Breakdown: Granular accuracy per signal type (e.g., Cooling Down: 49%, Extreme Surge: 84%, Stabilisation: 51%) — the most honest view of what works
• Pool-Level Blacklisting: Pools with persistently poor accuracy are automatically disabled. MAE kill switches remove pools with error rates exceeding 15%
• Bayesian Reliability Weighting: Engines adjust confidence based on each pool's historical accuracy and mean absolute error. Consistently predictable pools receive amplified signals; volatile pools are dampened
• Public Audit Trails: Complete prediction history with success/failure outcomes accessible via individual engine pages and the diagnostic report

Predictive Dashboard & Navigation

The prediction system comprises six interlinked pages forming a unified analytics silo:

• Predictive Dashboard: Meta-analysis combining all engines with system-wide accuracy windows, prediction type performance, and recent verifications
• Signal Confluence: Multi-engine trading signals with per-recommendation accuracy tracking
• APY Momentum: Live mean-reversion signals with Z-score, RSI, Bollinger Band, and EMA indicators
• TVL Momentum: Capital flow tracking with stabilisation predictions and market-relative scoring
• Surge Predictor: Ultra-selective extreme utilisation alerts — the system's highest-accuracy engine
• Reliability Index: Elite pool rankings based on consistency, stability, and current yield

DeFiStar.io is led by AC (@MooniTooki) who holds a BSc (Hons) in Computing and is a current Member of the Institution of Engineering and Technology (MIET). A systems engineer and data analyst with 20+ years industry experience now focused on yield optimising strategies and protocol security. With a background in decentralised architecture and real-time data systems, the goal going forward is to build tools that provide professional-grade intelligence to the wider DeFi community. MooniTooki's identity has been fully verified. Login to the Global ID USBC app to view profile and verification status.

Data Sources

• Blockchain Data: Direct RPC calls to protocol smart contracts

Data Quality Commitments

• No Manual Overrides: All rankings purely algorithmic
• No Commercial Bias: Zero listing fees or protocol partnerships
• Real-Time Verification: Users can verify data via protocol links
• Error Reporting: Contact us immediately if you spot discrepancies
• Methodology Documentation: Full technical specifications published

DeFiStar.io
Business Address: Clyde Offices, 2nd Floor, 48 West George Street, Glasgow, G2 1BP, Scotland, United Kingdom
Contact: Email
X (Twitter): @defistario X (Twitter): @moonitooki

For data discrepancies, technical issues, or general inquiries, please contact us via email. We aim to respond within 48 hours.

Legal & Regulatory:
DeFiStar.io operates as an independent data analytics platform. We are not authorised or regulated by the Financial Conduct Authority (FCA). We do not provide financial advice, investment services, or facilitate transactions. For full terms, see our Terms & Conditions.